Incorrect image size, relion 3.1

[yixiaozhang123]

Describe your problem

Hi,
I have four datasets from three different microscopes. I ran four polishing independently in the same relion folder and joined the shiny.star files in relion. When I used the joined star file to run refinement, it reported the incorrect image size error.

• RELION version [e.g. RELION-3.1-beta-commit-a6aaa5 (please see the title bar of the GUI)]

Error message:

fn_img= 9@Polish/job066/movies1/19may22i_grid4_00017gr_00022sq_v03_00006hln_00015enn_frames_shiny.mrcs XSIZE(img())= 280 YSIZE(img())= 280
 while XSIZE(exp_imagedata)= 266 and YSIZE(exp_imagedata)= 266
in: /programs/x86_64-linux/relion/3.1-beta_cu7.5/src/ml_optimiser.cpp, line 9051
ERROR: 
MlOptimiser::getMetaAndImageDataSubset ERROR: incorrect image size
master encountered error: === Backtrace  ===
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi(_ZN11RelionErrorC1ERKSsS1_l+0x41) [0x451721]
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi(_ZN11MlOptimiser25getMetaAndImageDataSubsetEllb+0x1f46) [0x630486]
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi(_ZN14MlOptimiserMpi11expectationEv+0xd98) [0x471658]
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi(_ZN14MlOptimiserMpi7iterateEv+0xc4) [0x4812e4]
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi(main+0x5f) [0x43b98f]
/usr/lib64/libc.so.6(__libc_start_main+0xf5) [0x7f39b6777b15]
/programs/x86_64-linux/relion/3.1-beta_cu7.5/bin/relion_refine_mpi() [0x43e181]
==================
ERROR: 
MlOptimiser::getMetaAndImageDataSubset ERROR: incorrect image size

==================

I paste the first lines of one subset of the joined star file.

version 30001

data_optics

loop_
_rlnOpticsGroupName #1
_rlnOpticsGroup #2
_rlnMtfFileName #3
_rlnMicrographOriginalPixelSize #4
_rlnVoltage #5
_rlnSphericalAberration #6
_rlnAmplitudeContrast #7
_rlnImagePixelSize #8
_rlnImageSize #9
_rlnImageDimensionality #10
_rlnCtfDataAreCtfPremultiplied #11
data1-krios3 1 mtf.star 1.045000 300.000000 2.700000 0.100000 1.045000 280 2 0
data2-krios1 2 mtf.star 1.070000 300.000000 2.700000 0.100000 1.070000 274 2 0
data3-krios1 3 mtf.star 1.070000 300.000000 2.700000 0.100000 1.070000 274 2 0
data4-krios2 4 mtf.star 1.096000 300.000000 0.001000 0.100000 1.096000 266 2 0

version 30001

data_particles

loop_
_rlnCoordinateX #1
_rlnCoordinateY #2
_rlnAutopickFigureOfMerit #3
_rlnImageName #4
_rlnMicrographName #5
_rlnOpticsGroup #6
_rlnCtfMaxResolution #7
_rlnCtfFigureOfMerit #8
_rlnDefocusU #9
_rlnDefocusV #10
_rlnDefocusAngle #11
_rlnCtfBfactor #12
_rlnCtfScalefactor #13
_rlnPhaseShift #14
_rlnAngleRot #15
_rlnAngleTilt #16
_rlnAnglePsi #17
_rlnOriginXAngst #18
_rlnOriginYAngst #19
_rlnClassNumber #20
_rlnNormCorrection #21
_rlnRandomSubset #22
_rlnLogLikeliContribution #23
_rlnMaxValueProbDistribution #24
_rlnNrOfSignificantSamples #25
1255.000000 1843.000000 0.516506 31@Polish/job068/movies2/19jul08c_b2g3_00020gr_00153sq_v02_00002hl_00006enn_frames_shiny.mrcs MotionCorr/job007/movies2/19jul08c_b2g3_00020gr_00153sq_v02_00002hl_00006enn_frames.mrc 2 7.098020 0.018632 25636.865234 25377.695312 160.883070 0.000000 1.000000 0.000000 -135.65107 61.817513 16.155556 0.898641 1.968641 1 1.167588 1 2.246716e+05 0.044261 47
3233.000000 2258.000000 0.709947 45@Polish/job066/movies1/19may22i_grid4_00011gr_00006sq_v03_00005hln_00016enn_frames_shiny.mrcs MotionCorr/job006/movies1/19may22i_grid4_00011gr_00006sq_v03_00005hln_00016enn_frames.mrc 1 3.714352 0.052656 25121.041563 24768.367580 -157.51869 0.000000 1.000000 0.000000 -131.55552 119.241584 -62.47346 -2.56581 -0.47581 1 1.153912 2 2.357244e+05 0.655179 11
151.000000 2341.000000 0.381428 53@Polish/job068/movies2/19jul08c_b1g3_00020gr_00015sq_v02_00002hl_00016enn_frames_shiny.mrcs MotionCorr/job007/movies2/19jul08c_b1g3_00020gr_00015sq_v02_00002hl_00016enn_frames.mrc 2 4.096572 0.049726 24387.126953 24209.390625 148.792200 0.000000 1.000000 0.000000 63.142522 75.667826 172.323983 -0.17136 -1.24136 1 1.161556 2 2.178788e+05 0.132492 41
1749.000000 3056.000000 0.450146 88@Polish/job069/movies3/19sep30h_Grid1_00010gr_00052sq_v02_00002hl_00015enn_frames_shiny.mrcs MotionCorr/job008/movies3/19sep30h_Grid1_00010gr_00052sq_v02_00002hl_00015enn_frames.mrc 3 4.425309 0.036512 28798.080788 27825.553158 55.585456 0.000000 1.000000 0.000000 118.437951 55.429976 -3.43885 1.366287 -0.77371 1 1.220891 1 2.149354e+05 0.065174 60
1372.000000 2398.000000 0.408276 66@Polish/job070/movies4/19oct15d_Grid3_00016gr_00058sq_v02_00002hln_00002enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3_00016gr_00058sq_v02_00002hln_00002enn_frames.mrc 4 6.126359 0.004945 31992.272671 27117.596299 92.159060 0.000000 1.000000 0.000000 86.783647 55.029820 -56.15928 1.144529 0.048529 1 1.237622 2 2.255628e+05 0.361023 18
634.000000 1327.000000 0.372096 19@Polish/job070/movies4/19oct15d_Grid3a_00021gr_00191sq_v02_00003hln_00007enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3a_00021gr_00191sq_v02_00003hln_00007enn_frames.mrc 4 3.583920 0.052278 22746.535651 20853.145731 88.592321 0.000000 1.000000 0.000000 -172.17202 104.169069 -67.92351 -2.02647 -4.21847 1 1.203617 1 2.108789e+05 0.336099 26
2981.000000 1920.000000 0.488336 58@Polish/job070/movies4/19oct15d_Grid3_00009gr_00038sq_v02_00002hln_00009enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3_00009gr_00038sq_v02_00002hln_00009enn_frames.mrc 4 6.074441 0.017843 28652.410629 26716.470662 -179.61637 0.000000 1.000000 0.000000 81.027849 34.078715 -106.01544 0.165527 -2.14347 1 1.131400 1 2.144561e+05 0.173949 22
112.000000 3044.000000 0.417037 90@Polish/job070/movies4/19oct15d_Grid3a_00022gr_00180sq_v03_00004hln_00011enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3a_00022gr_00180sq_v03_00004hln_00011enn_frames.mrc 4 5.973200 0.033235 24573.230130 21686.866020 -178.94190 0.000000 1.000000 0.000000 104.649600 123.590662 76.984276 1.144529 0.165527 1 1.076793 1 2.257674e+05 0.105919 73
2375.000000 1399.000000 0.370700 42@Polish/job070/movies4/19oct15d_Grid3a_00016gr_00152sq_v02_00003hln_00011enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3a_00016gr_00152sq_v02_00003hln_00011enn_frames.mrc 4 8.533143 0.001025 22978.833754 18533.780550 90.842861 0.000000 1.000000 0.000000 -127.86883 46.781757 149.869776 5.645527 -0.93047 1 1.156407 1 2.082427e+05 0.191841 30
2748.000000 2612.000000 0.493394 67@Polish/job070/movies4/19oct15d_Grid3_00022gr_00059sq_v02_00002hln_00007enn_frames_shiny.mrcs MotionCorr/job009/movies4/19oct15d_Grid3_00022gr_00059sq_v02_00002hln_00007enn_frames.mrc 4 9.431368 -0.01025 28052.530664 24690.788293 94.007723 0.000000 1.000000 0.000000 129.972721 27.502984 -97.54721 1.144529 5.645527 1 1.154269 2 2.182097e+05 0.164259 31
3058.000000 2288.000000 0.647688 42@Polish/job066/movies1/19may22i_grid4_00012gr_00008sq_v02_00006hln_00014enn_frames_shiny.mrcs MotionCorr/job006/movies1/19may22i_grid4_00012gr_00008sq_v02_00006hln_00014enn_frames.mrc 1 4.685425 0.035603 30219.920992 29701.870138 -158.45377 0.000000 1.000000 0.000000 -122.31310 140.973199 -60.15296 2.659189 1.614189 1 1.197359 2 2.354910e+05 0.447865 4

Thanks,
Yixiao

[biochem-fan]

(I fixed your markup)

This seems related to #519.

Does this happen when you disable preread particles to RAM and copy particles to scratch?

[yixiaozhang123]

No, I comfirmed that the particles were pre-read to RAM.

[biochem-fan]

Please test if it happens even without preread and scratch.

[yixiaozhang123]

It reported the same error if I chose no-preread and no scratch, even with no gpu.
I also tried the star files joined from Extract or CtfRefine, still got the same error.
HOWEVER, if the particles were extracted in a same box size (280 pixels) with their different pixel size, the joined particles can be refined with no error if the reference map has a box size of 280 pixels (can be in different pixel size, but must be in the same box size of 280). It looks like different optic groups should have a same box size in the relion version I'm using.

[biochem-fan]

OK, I finally found a reproducible case.

Meanwhile, if you set Use parallel disc I/O: Yes, the problem should disappear.

[biochem-fan]

@scheres The problem is that we are resizing exp_imagadata to mymodel.ori_size in https://github.com/3dem/relion/blob/ver3.1/src/ml_optimiser.cpp#L8980.

[yixiaozhang123]

It works with "Use parallel disc I/O: Yes" setting. Thanks!
Do you have any recommendation on which reference map we should use for data combination? Largest pixel size? Smallest box width?

[biochem-fan]

Largest pixel size? Smallest box width?

It does not matter really but choosing the largest pixel size and the smallest box size is the most efficient.

[davidmbelnap]

I submitted #519, which is the same error, and I was told to continue the discussion here. Two things:

1. I was asked if I had the problem if "preread particles to RAM" and "copy particles to scratch" were disabled. Yes, I had the problem if those two options were not used.

2. I was told to "use parallel disc I/O". If I use that option, the problem does not occur.

[davidmbelnap]

Thank you for working on the problem.

[biochem-fan]

Using different box sizes with Parallel disk I/O: No turned out to be non trivial. We decided not to support it. Please use Parallel disk I/O: Yes when you are merging particles with different box sizes.

We added check for this (e023ce5).